1、Data Center
RLP:请求级并行,松耦合
2、Cloud Computing
- Economic of Scale
- Supported by Data Centers
Issues:
(1)Multi-Tenant Isolation
- Resource Allocation:CPU、Memory、IO devices、Network
- Application Isolation:
(2)Load Balancing
- Traffic Pattern
- Solution:
Background Jobs
Use the resource off-peak
Combine different types of users
(3)Energy Efficiency
(4)Dependability
3、Distribution Systems
- Map Reduce
- Google File System(GFS)
分布存储的问题:
- Data Deduplication:避免重复存储
- Erasure Codes for Storage:Provides Redundancy,Recovery Costs
一致性问题
- Sever can
- CAP Theorem:consistency,availability、Partition tolerance三者权衡
分布式共识(consensus)
- The Paxos Protocol
- Distributed Locks Service